Processing terrestrial imagery


I am just getting familiar with ODM and I am wondering if it can allow for processing terrestrial imagery to obtain 3D products such as DEM or point cloud. Does anyone have any experience/tips reagrding this type of processing?

In my case, I am interested in using frames from a dashcam videos to reconstruct the road geometry. At the moment I do not have GPS data associated to each frame but I can probably find a way to geotag the images.

Thanks for your answers.

1 Like

Absolutely can!

Klas does some amazing stuff with a GoPro and his feet :slight_smile:

We even have a few folks here that have done dashcam reconstruction, though I have no links of final products.

I’ve done walk-arounds inside my apartment, outside, etc and they work well. Sometimes you’ll need to constrain the reconstruction with --auto-boundary or --boundary if you’re getting a lot of sky/background, but in my experience, it works well.


Fantastic, thanks for the answer and the link!

are there any important parameters to set or think of other than --auto-boundary?

1 Like

I’m not sure, to be honest! With my mobile phone, I have needed to use --feature-quality ultra to get reliable reconstructions, but I’m not sure that’d be necessary with a less-noisy sensor like if I used my Fuji X-T2 (which I have yet to try).

I would suggest tackling a really small sample reconstruction to get a feel for your camera and scenes and see how it goes. You can always share it here and ask for help so folks can give their input :slight_smile:


tanks for the tips, I will try to get the images geotagged and then I will upload a sample here.


This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.